The Role of Gas in Determining Image Quality andResolution During In Situ Scanning Transmission ElectronMicroscopy ExperimentsYuanyuan Zhu*[a] and Nigel D. Browning[a, b]

Introduction

Directly monitoring heterogeneous catalysts under catalytic

conditions is key to obtaining insights into their behavior. Cur-rently, environmental (scanning) transmission electron micros-

copy (E(S)TEM) is the only characterization tool that can pro-vide atomic-site specific observations of local catalyst surface

structure during solid–gas reactions.[1] In this approach, a

volume of pressurized gas is introduced and maintained tightlyaround a TEM sample through either a differential pumping

mechanism (DP) or by a membrane windowed (MW) gas cell.[2]

The use of in situ TEM has progressed from pioneering early

work featuring self-built modifications to existing microscopes/holders[3] to the use of commercially manufactured[4] instru-ments. However, the gaseous environment (and its container)

continues to limit the (S)TEM sensitivity and resolvability incharacterizing the solid catalyst owing to an increase in scat-tering events from the background gases/windows. To achievebetter imaging sensitivity and resolvability in catalytically rele-

vant gaseous environments, it is therefore crucial to under-

stand how the gas molecules affect the scattering of fast elec-

trons contributing to the atomic-scale image.[5]

There are two basic atomic imaging modes in (S)TEM, the

conventional high-resolution transmission electron microscope(HR-TEM) phase contrast image and the incoherent annular

dark-field scanning transmission electron microscopy (ADF-

STEM) image. Although environmental HR-TEM (HR-ETEMmode) has been widely used for in situ observations of hetero-

geneous catalysts and provides quantitative atomic informa-tion with the aid of image simulations,[6] environmental ADF-

STEM (ESTEM mode) has already started demonstrating prom-ising advantages for imaging supported metallic nanocata-lysts[7] with single-atom sensitivity,[8] and for offering directly in-

terpretable images of atomic-scale dynamics in complex oxidecatalysts.[9] Additionally, analytical techniques including energy-dispersive X-ray spectroscopy (EDX) and electron energy lossspectroscopy (EELS) in ADF-STEM mode (i.e. , STEM-EDS and

STEM-EELS) can also provide localized compositional and elec-tronic information.[7b, 10] Despite these potential advantages of

ESTEM, the effect of the gas environment on the image qualityhas so far only been discussed for the HR-ETEM imagingmode.[5a, 11]

In this work, we conduct the first experimental evaluation ofthe quality and resolution of the DP-ESTEM imaging mode (in

a non-aberration-corrected STEM) under the influence of vari-ous gas environments at different pressures. One of the main

constrains of DP-ESTEM is the lower DP aperture, which limits

the ADF detector outer acceptance angle. We examine thisissue by using a uniform solid sample—the standard lacy

amorphous carbon, whose low atomic weight (Z) renders it asa touchstone for evaluating the gas effects on ESTEM contrast

and on image signal-to-noise ratio (SNR), with different ADFcollection ranges. As catalysts and particularly their atomic de-

As gas–solid heterogeneous catalytic reactions are molecular

in nature, a full mechanistic understanding of the process re-

quires atomic-scale characterization under realistic operatingconditions. Although atomic resolution imaging has become

routine in modern high-vacuum (scanning) transmission elec-tron microscopy ((S)TEM), both image quality and resolution

nominally degrade on introduction of the reaction gases. Inthis work, we systematically assess the effects of different

gases at various pressures on the quality and resolution of

images obtained at room temperature in the annular dark field

STEM imaging mode by using a differentially pumped (DP) gas

cell. This imaging mode is largely free from inelastic scatteringeffects induced by the presence of gases and retains good

imaging properties over a wide range of gas mass/pressures.We demonstrate the application of ESTEM with atomic resolu-

tion images for a complex oxide alkane oxidation catalystMoVNbTeOx (M1) immersed in light and heavy gas environ-ments.

[a] Dr. Y. Zhu, Prof. Dr. N. D. BrowningPhysical & Computational Science DirectoratePacific Northwest National LaboratoryRichland, WA 99352 (USA)E-mail : yuanyuanzhu2014@gmail.com

nigel.browning@pnnl.gov

[b] Prof. Dr. N. D. BrowningDepartment of Materials Science and EngineeringUniversity of WashingtonSeattle, WA 98195 (USA)

Supporting information and the ORCID identification number(s) for theauthor(s) of this article can be found under https://doi.org/10.1002/cctc.201700474.

This publication is part of a Special Issue on “Advanced Microscopy andSpectroscopy for Catalysis”.

ChemCatChem 2017, 9, 3478 – 3485 T 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim3478

Full PapersDOI: 10.1002/cctc.201700474

tails in gas environments are usually the main interest ofin situ microscopy observations, we also show how the gases

affect the ESTEM SNR and resolution in atomic resolutionimages of crystalline MoVNbTeOx (M1) catalysts. All evaluations

were conducted at room temperature, for high temperatureimaging also depends on the stability of the heating stage

used, which is beyond the scope of this study.

Results and Discussion

Effect of gas on ESTEM image quality for amorphous carbon(i.e. , at low spatial resolution)

Figure 1 a–d shows experimental DP-ESTEM images of regions

with and without the solid carbon sample, in four commonlyused gases at pressures from high vacuum (2 V 10@7 mbar) to

10 mbar.

Considering that the DP gas cell apertures (especially theones below the sample) confine the scattering angle of exit

electrons and impose a 71 mrad cut-off, we choose an ADF de-tector inner semi-angle of approximately 21 mrad to provide

an adequate collection range of 21–71 mrad (i.e. , a low-angleannular dark field (LAADF) with intensity ~Za, a : 1.0–1.3[12]).Notably, a low probe current of <3 pA was used throughout

this work to reflect the practical ESTEM conditions usually usedfor imaging beam-sensitive catalysts (for dose and dose-ratedefinitions see the Experimental Section). As shown in, for ex-

ample, the intensity line profiles of the pressure series in N2

gas (Figure 1 e), the gas environment leads to an increasing in-

tensity of the gas background (and of the solid sample as thegas is all around the sample) ; meanwhile, the noise level in the

linescans also increases slightly along with N2 pressure (sec-tion 1 in the Supporting Information).

More importantly, if the N2 pressure series is normalized tothe corresponding gas background intensity (Igas), in Figure 1 f,the contrast of the carbon decreases as the pressure increases.

We define the ESTEM contrast of the lacy carbon as the differ-ence between the intensity of carbon and gas backgroundverses the gas intensity (Weber contrast): (Icarbon@Igas)/Igas. ThisESTEM carbon contrast, if normalized to the initial contrast in

the vacuum (Figure 1 g) decreases by about 55 % in N2 and Arat 10 mbar. The “useful signal” from a sample degrades as the

gas pressure/mass increases. To examine to what extent the

presence of gases influences the “image signal of a sample”,here, we define the signal/noise ratio (SNR) as the intensity

from the carbon (Icarbon@Igas) divided by the standard deviationof the sample raw counts DIcarbon. Thus, the higher the SNR

(the values in Figure 1 h at low gas pressures), the better theimaging ability to observe the structural and chemical features

of a sample. In contrast, as the gas pressure/mass increases

the net carbon intensity starts to become comparable with itsnoise level, inducing uncertainties in the observation of the

solid sample. The SNR is found to be more strongly confinedby the gas environment if the ESTEM images are collected

Figure 1. Experimental ESTEM images of gas background and amorphous lacy carbon from high vacuum to 10 mbar in a) He, b) N2, c) H2, and d) Ar. Allimages are under the same magnification, and the scale bar is 200 nm. e) Intensity line profiles in raw counts from gas background to the carbon sample inN2 at different pressures. The measurement of Igas, Icarbon, and DIcarbon are defined. f) Normalized experimental carbon intensity to corresponding Igas, showingthe carbon contrast change against the N2 background at different pressures. The calculations of Contrastcarbon and SNRcarbon are defined. g) Normalized experi-mental ESTEM Contrastcarbon ((Icarbon@Igas)/Igas) and h) SNRcarbon in He, H2, N2, and Ar gas environments as a function of pressure. The Rose criterion (SNR>5) ismarked to show the preserved high SNR of carbon in heavy gas up to 10 mbar.

ChemCatChem 2017, 9, 3478 – 3485 www.chemcatchem.org T 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim3479

Full Papers

under a medium-angle annular (MAADF) setting (section 2 inthe Supporting Information). However, for the LAADF mode

with a relatively large collection range, the ESTEM SNR of thelight carbon remains above the Rose criterion[13] in 10 mbar Ar.

Thus, for a given catalytic sample (often a fixed Z-number) theinner ADF collection angle can be used to leverage heavy gas

and high pressure environment with controlled electron dose/dose-rate in DP-ESTEM, without introducing contrast reversals

(even for light carbon) while maintaining a good sample con-

trast and SNR.Considering that the STEM ADF detector is an electron-

counting device, the raw image intensity faithfully counts theforward scattered electrons collected by the detector. The ob-

served increases in ESTEM intensity and more importantly inthe noise level from the area of the carbon suggest that thereare additional “unwanted” electrons scattered by the gas mole-

cules onto the ADF detector. These “extra electrons” cause thechanges in the intensity as well as in the noise level in environ-mental images. To quantitatively assess to what extent thisgas-induced scattering affects the ETEM and ESTEM imaging

modes, we calculated the cross sections of elastic and inelasticscattering by the gases under the two imaging settings (sec-

tion 3 in the Supporting Information). As shown in Figure 2,

the ETEM imaging mode records almost all the extra inelasticnoise from the gases at low scattering angles (open symbols);

whereas the ESTEM mode excludes most of it (inelastic charac-teristic scattering angles of typical gases are <0.25 mrad in

Table S1 in the Supporting Information). In the ESTEM mode,the relatively low total extra scattering (from the gas mole-

cules) explains the high imaging sensitivity (i.e. , SNR) in thesample. The advantage of the ADF-STEM mode in revealing

catalytic nanoparticles on a thick support background[14] also

applies to imaging the materials in gases (the gas volume andthe windows for MW gas cells are now the background of thesignal).

Effect of gas on ESTEM atomic resolution image quality forthe M1 catalyst

Next, we investigated the gas effect on the atomically resolved

ESTEM images. Because of its large lattice parameter and goodstability in a gas environment,[9] the crystalline M1 catalyst was

employed for this evaluation. Figure 3 a, b presents high-resolu-tion ESTEM pressure series of the M1 crystal in [0 0 1] projec-

tion in He and N2 environments, respectively. To further quanti-fy the gas effect, we performed a fast Fourier transform (FFT)based signal analysis, which is better suited for measuring

atomically resolved lattice images. For example, Figure 3 cshows the FFT-based SNR measurement on an M1 atomicESTEM image (details of this method are given in the Experi-mental Section). The results show, in Figure 3 d, the SNRM1 is,

as expected, higher in the light He than that in the heavier N2

at the same pressure. At 10 mbar, the ESTEM SNRM1 drops to

60.6 % in He, and to 42.0 % in N2 under the low beam current

condition in this work. In the case of the He environment, asthe ESTEM resolution hardly decreases up to 10 mbar, the deg-

radation of the SNRM1 in He must be mainly caused by postspecimen scattering of the image-forming electrons. That is to

say, as the electrons carrying the sample information try toreach the ADF detector, they could be further scattered by the

gas molecules in the post-sample gas pathway,[5a] and are re-

directed to a final scattering angle that is outside of the detec-tor collection range, leading to a loss of signal from the solid

sample. For the gases heavier than He, both pre-sample andpost-sample gas scatterings may contribute to the degradation

in ESTEM SNRsample.

Figure 2. Theoretically calculated cross sections of elastically and inelasticallyscattered electrons at different scattering angular ranges as a function ofthe atomic number of the object the electrons (with an incident energy of300 kV) interact with. Different types of gases are marked based on theiratomic number. For a Titan DP environmental microscope, the scatteredelectrons collected in the ETEM imaging mode is 0–71 mrad (assume no ob-jective aperture used), and an annular range of 21–71 mrad was collectedby the LAADF-ESTEM imaging mode used in this work.

Figure 3. Experimental high-resolution ESTEM images of the M1 catalyst in [0 0 1] zone axis from high vacuum to 10 mbar in a) He and b) N2. Scale bar is10 nm. c) The quantification of atomic ESTEM image SNR of M1 by using the fast Fourier transform (FFT) measurement (for details see the Experimental Sec-tion). d) The FFT measured SNR of the M1 experimental ESTEM images in He and N2 gas environments as a function of pressure.

ChemCatChem 2017, 9, 3478 – 3485 www.chemcatchem.org T 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim3480

Full Papers

Figure 4 a, b present the approach and results of the ESTEMresolution evaluation over the M1 crystal in He and in N2 up to

10 mbar. The large unit cell and thus densely packed diffrac-tion spots of the M1 crystal in the [0 0 1] projection provide

high sensitivity for resolution estimation. By measuring theESTEM resolution on multiple M1 particles at different gas

pressures (section 6 in the Supporting Information), we findthat the He gas hardly affected the ESTEM resolution (resolu-

tion drop <0.02 a at 10 mbar) whereas the presence of N2

leads to noticeable resolution degradation (>0.40 a at10 mbar), under the low beam current imaging condition. To

understand this difference in the ESTEM resolution depend-ence on the gas type, we studied the cause of STEM resolution

degradation. It is known that the resolution of STEM is gov-erned by the size of the converged probe.[15] In conventional

microscopes, the integrity of the STEM probe is preservedowing to the large electron mean free path under highvacuum; in the DP gas cell, prior to interacting with the

sample, the STEM probe has to pass through the pre-samplegas species, which leads to extra scattering and possible diffu-

sion of the originally sharp probe. If we assume that the gasspecies along the pre-sample pathway are acting collectivelyas a solid foil, the gas-induced STEM probe spreading can thenbe considered in a similar way to the beam broadening b

caused by a (thick) sample. This probe broadening b can be es-timated quantitatively based on Bothe’s multiple-scatteringtheory,[16]

b ¼ 1:05> 103ð 1WÞ1=2 Z

E1þ E=E0

1þ E=2E0

L3=2 ð1Þ

in which 1, W, and Z are mass density, the mean molar mass,and the mean atomic number of a chosen solid, E is the micro-

scope acceleration voltage in eV and E0 is 1 eV. It shows that bincreases with the 3/2 power of the sample thickness.

Considering that silicon nitride (Si3N4) is commonly used as

the window material for the membrane windowed (MW) envi-ronmental gas cell, we chose it as the gas-equivalent solid foil.

Based on the idea gas law, at a given pressure (P) and temper-

ature (T), the reversed molar volume of a gas with a thicknessof Lgas is equal to the reversed molar volume of Si3N4 with a

thickness of LSi3N4,

PRT

Lgas ¼1

WLSi3N4 ð2Þ

in which R is the gas constant, W is 20.0 g mol@1, and 1 is 3.2 V106 g m@3 for Si3N4. However, the above equation does not

specify the type of the gas, that is, any 10 mbar gases at 300 Kwith a gas pathway of 2.7 mm are equivalent to 6.75 nm Si3N4.

Clearly, this simple estimation is too generalized for any

10 mbar gases regardless of their type and thus cannot explainthe difference in ESTEM resolution observed in He and in N2.

To take the nature of the gas into consideration, we revise theabove equation by weighting both sides with the total scatter-

ing cross section s and distinguishing gases by including thenumber of atoms in a gas molecule M,

MP

RTLgassgas ¼

1

WLSi3N4sSi3N4 ð3Þ

With this modification, the equivalent Si3N4 thicknesses ofHe and N2 are 0.51 nm and 21.29 nm, respectively at 10 mbar

(Table S2 in the Supporting Information). Thus, the theoreticalprobe broadening b as well as the corresponding ESTEM reso-

lution reduction DR were estimated. The theoretically estimat-ed ESTEM resolution drops are plotted as solid lines in Fig-

Figure 4. a) Experimental ESTEM resolution estimation based on the FFT ofan atomic M1 image. b) Theoretical and experimental ESTEM resolution ofM1 in He and N2 up to 10 mbar. c) Schematic illustration of pre-sample gasintroduced STEM probe spreading.

ChemCatChem 2017, 9, 3478 – 3485 www.chemcatchem.org T 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim3481

Full Papers

ure 4 b. It shows that at 10 mbar, ESTEM resolution change isnegligible in the He gas; whereas 2.7 mm N2 leads to about

0.25 a degradation in ESTEM resolution in theory. Notably theapparent probe broadening b should be added to the original

probe size in quadrature to reflect the Gaussian-distributionnature of the STEM probe intensity.

Thus, as schematically illustrated in Figure 4 c, after takingthe gas type into account, the above theoretical estimation ex-

plains qualitatively the observed difference in the trend of the

ESTEM resolution up to a higher pressure between the lightand heavy gas environment. It is worth mentioning that the

unpromising ESTEM resolution measured in N2 (blue trianglesin Figure 4 b) was under the low beam current imaging condi-

tion used throughout this work. This low-dose resolution canbe boosted by adjusting the STEM imaging conditions if the

sample is not damaged and there are no kinetic changes in

the reaction caused by the elevated current (details will be dis-cussed in the next section).

Comments on electron dose and dose rate

To examine if the low beam current imaging conditions usedin this work become limiting factors for ESTEM image quality

and resolution with the presence of gas, we tested the aboveESTEM SNRs, contrast, and resolution at different electron dose

and dose rates, respectively, in 10 mbar N2 environment. Asshown in Figure 5, both of the SNRs of the amorphous carbon

and of the M1 sample, and the ESTEM resolution increase

along with the increase of total accumulated electron dose perimage frame; whereas, such correlations are not observed on

varying the dose rate (Figure S7 in the Supporting Informa-tion). In particular, for the atomically resolved ESTEM images of

the M1 catalyst in 10 mbar N2, by increasing the electron dosefrom the pre-set low dose (8.24 V 102 e a@2) to a relatively

higher dose of 6.57 V 103 e a@2, the SNRM1 is increases from 7.21

to 15.20 (the orange square in Figure 3 d), and the resolutiondrop is mitigated to about 0.20 a (Figure 4 b), agreeing well

with the theoretical resolution estimation. This suggests thatboth ESTEM SNR and resolution depend on the accumulated

electron dose in an image, which is fundamentally differentfrom the dose-rate dependence found for the ETEM resolutio-

n.[11a, b] On the other hand, the ESTEM contrast of the amor-phous carbon remains largely constant regardless of the elec-

tron dose and dose rate used for imaging (notably the fluctua-tions in contrast at extremely low dose/dose-rate are causedby the high uncertainties in determining the contrast). Thissuggests that the “Z-contrast” nature of the STEM imagingmode, and thus the elemental indications of the STEM intensi-

ty, is still valid for ESTEM imaging with the consideration forenvironmental contribution.

Unlike the convoluted relationship between the TEM beamand gas species,[5] a relatively simple picture might be possibleto elucidate the gas effects in the STEM imaging mode. One

such attempt is schematically illustrated in Figure 6 a. In thepost-sample region at the bottom half of the gas-cell chamber,

the STEM image-forming electrons carrying the information ofthe sample are generally inelastic in nature. The interactions

between these electrons (i.e. , sample signal) and the gas spe-cies (and/or bottom silicon nitride window) mainly lead to a

diffusion of the STEM signal by scattering (some of) the image-forming electrons off the finite ADF detector. Thus, it results in

losses in SNR and in resolution owing to the lack of signal. By

increasing the electron dose through a longer dwell time, forexample, the signal level can be retained and then recoversthe ESTEM SNR and resolution. In other words, if a catalystsample can tolerate an accumulated electron dose, the effects

of gas species in the post-sample region can be largely allevi-ated. For example, excellent atomic ESTEM SNR was observed

by using an electron dose of 104 e a@2.[7b]

With the ESTEM signal diffusion retrieved by accumulatingmore electron doses, we can then propose the theoretically

obtainable ESTEM image resolution for both DP-ESTEM andMW-ESTEM. Based on the multiple-scattering theory, in Fig-

ure 6 b, the estimated resolutions for original STEM probe sizesof 0.78 a (e.g. , JEM-ARM 200F, probe-corrected), of 1.36 a (e.g. ,

Titan G2 300 kV), and of 1.90 a (e.g. , Tecnai G2 F30 S-Twin) are

plotted as a function of the thickness of the gas-equivalent sili-con nitride (window). In general, the ESTEM probe broadening

accelerates as the gas mass/pressure (or the pre-samplewindow thickness) becomes greater. For the DP gas cell (used

in this work), the effect of 10 mbar N2 on the ESTEM resolutionis similar to the best obtainable resolution of a MW gas cell

Figure 5. Experimentally measured electron dose dependent of a) theESTEM SNR and contrast of carbon, and b) the SNR and ESTEM resolution(Figure S6 in the Supporting Information) over the M1 catalyst, in 10 mbarN2.

ChemCatChem 2017, 9, 3478 – 3485 www.chemcatchem.org T 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim3482

Full Papers

with a 20 nm-thick pre-sample Si3N4 window (if assuming the

sample is attached to the top inside of the upper window). Ina similar sense, the upper limit of 20 mbar N2 in a DP gas cell

(of 2.7 mm pre-sample gas pathway) is equivalent to a MW cellof an approximately 43 nm Si3N4 window (not shown here). Onthe other hand, as indicated by the comparison between a10 nm and a 50 nm Si3N4 pre-sample window, the achievable

ESTEM resolution for MW gas cells depends strongly on thewindow thickness (and for that matter, also the sample posi-tion). To eliminate the probe broadening effects, a Si3N4

window thinner than <6 nm (or the N2 in 2.7 mm gas-pathwayis lower than approximately 3 mbar) is required. For a practical

ESTEM probe of <2.0 a applicable for most metallic and metaloxide catalysts, an advanced objective lens is encouraged for

DP-ESTEM and probe correction is required for MW-ESTEM.

Conclusions

Catalysis has benefited greatly from atomic-level imaging in

modern (S)TEM, and in turn the kinetic nature of catalysis hasalso shaped the development of new capabilities in (S)TEM.

The recent advancements in situ environmental (S)TEM havefurther opened up unprecedented opportunities for obtainingexperimental insights into the working structure of heteroge-neous catalysts in technologically relevant environments. In ad-dition, the presence of a catalytic environment such as pressur-ized gases poses new challenges to the (S)TEM sensitivity and

resolvability and to our understanding of gas–electron interac-tions. In this work, we investigated both experimentally and

theoretically the gas environment effects on image quality (interms of sample contrast and SNR) and resolution in the direct-ly interpretable ADF-STEM imaging mode in DP-ESTEM. Theconclusions and guidelines identified are as follows:

i) The preconceived constrain imposed by the lower gas

aperture in DP-ESTEM can be alleviated by loosening the innerADF collection angle, without introducing image contrast re-

versals.

ii) The absence of the extra inelastic noise owing to the an-nular collection mode of the ADF-STEM imaging allows the

ESTEM imaging mode to be free of intensity loss, and to main-tain a good sensitivity in imaging the solid sample (sample

SNR) with the presence of gases. Unlike the ETEM mode,ESTEM does not require a monochromator to achieve high

quality imaging.

iii) The ESTEM sample contrast, owing to the incoherentnature of the ADF-STEM imaging mode, remains correlated to

the (mean) atomic number of the solid sample with the con-sideration of the elastic contribution from the gas environ-

ment.iv) The obtainable ESTEM resolution is determined mainly

by the probe broadening effects induced by the pre-sample

gas and electron beam interactions. Our theoretical estimationon the resolution shows that a probe-corrector is a necessity

for high-resolution ESTEM imaging.v) The practical ESTEM sample SNR and resolution are elec-

tron dose dependent, owing to the signal diffusion effectcaused by the gas species in the post-sample region. In prac-

tice, the fact that the SNR and resolution both increase along

with the accumulation of electron dose is beneficial for achiev-ing high sensitivity and at the same time high-resolution

ESTEM observations.The ultimate challenge in imaging working catalysts in gas–

solid heterogeneous catalysis is to be able to directly visualizethe gas adsorbent at a catalyst surface with the reaction raterelevant high temporal resolution. Phase contrast in conven-

tional HR-TEM has been used to image light materials (i.e. ,weak-phase objects) with the additional help of either phase

plate[17] or using off-axis holography.[18] Although the frametime for one TEM image is relatively short compared with a

typical STEM image, a HR-TEM focal-series is required foratomic interpretation, which takes a much longer time and

then is low in temporal resolution. Recent developments in ad-

vanced STEM imaging have begun to leverage light-elementimaging by, for example, ptychography,[19] and to accelerate

scanning speed without compromising image SNR by novelscan sampling.[20] The combinations of these recent develop-

ments with the current high quality atomic ESTEM discussed in

Figure 6. a) Schematic illustration of the pre- and post-sample gas effects onESTEM probe and signal. b) Theoretical estimated ESTEM resolution as afunction of silicon nitride thickness (Table S3 in the Supporting Information).

ChemCatChem 2017, 9, 3478 – 3485 www.chemcatchem.org T 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim3483

Full Papers

this work, makes it a promising in situ tool-box for investigat-ing catalytic dynamics.

Experimental Section

Sample preparation

Standard TEM lacey carbon Cu grids (Ted-Pella) were used as theywere for ESTEM imaging of amorphous carbon. High purity crystal-line MoVTeNb mixed oxide catalyst (M1 phase) was obtained byhydrothermal synthesis.[9] TEM samples were prepared by crushingthe M1 catalyst into a fine powder, and dry loading on an ultrathincarbon film on the lacey carbon support film Au grids (Ted-Pella).M1 samples were plasma cleaned 10 to 15 s prior to ESTEM imag-ing.

Environmental scanning transmission electron microscopyimaging

We employed a dedicated ETEM (FEI Titan 80/300) with differential-ly pumped gas cell system (objective lens pole pieces separation~5.4 mm) housed in the EMSL user facility at PNNL. All ESTEMimages shown in this work were acquired at 300 kV with a conver-gence semi-angle of 9.9 mrad without probe aberration correction.A low-angle annular dark field (LAADF) STEM mode with a collec-tion range of 21–71 mrad (upper collection angle is limited by theDP gas cell aperture) was used throughout this work. A higherADF detector inner semi-angle of 48 mrad (MAADF-STEM) was alsotested with the presence of gases and details on the justificationof using the LAADF-ESTEM imaging mode are given in section 2 inthe Supporting Information. This microscope equipped with afield-emission gun produces a stable and bright electron probe. Todetermine the accurate electron current on the sample, we con-ducted a Faraday cup measurement by using an analytical holder(Gatan, Inc.) to calibrate the microscope screen dose-mete espe-cially for the low current density region of <3 pA. To reproducethe low electron dose-rate and low accumulated dose imagingconditions that are often used for imaging delicate catalysts, a lowbeam current was kept around 1.5–3 pA and a short dwell time of4 ms/pixel was used for collecting all the ESTEM images (unlessstated otherwise). The image magnification (and hence dose-rate)was chosen based on the size of the sample feature imaged. Inparticular, for the amorphous carbon, we used a magnification ofV 115 k corresponding to 3.34 a/pixel, dose rate 0.26 e a@2 s@1 anddose per frame 4.50 e a@2 ; the common magnification used foratomically resolved M1 catalyst is V 1800 k corresponding to0.22 a/pixel, dose rate 4.79 V 101 e a@2 s@1, and dose per frame8.24 V 102 e a@2 (unless stated otherwise). The total dose was keptunder the damage threshold for each M1 particle imaged.[9]

Contamination from hydrocarbon build up in the ETEM columncan impose difficulties during STEM imaging especially for a multi-user instrument. In this work, we found that the built-in plasmacleaner in the ETEM helped alleviate this issue. Prior to each ESTEMsession, the plasma cleaning procedure was performed at 14 W(energy) for 10 h followed by 4 h purging with nitrogen. Externalplasma cleaning of the TEM sample is also encouraged to furthereliminate contamination. During the ESTEM experiment, the back-pressure around the sample was pumped down to <2 V 10@7 mbarafter inserting the sample holder to set up a clean background.Ultra-pure He, H2, N2, and Ar was introduced around the sample to0.1, 1, 5, and 10(:5 %) mbar, and flowed for about 10 min everytime after reaching the targeting pressure before imaging. The op-

timal STEM imaging defocus was achieved by maximizing the con-trast.

ESTEM image quantification

It is essential to read out the raw electron counts from an ESTEMimage instead of the apparent image intensity, which varies de-pending on the brightness/contrast (B/C) setting. To avoid this, weused the original 16-bit.ser files or converting them to 16-bit (not8-bit).tiff image files (so they can be read in DigitalMicrograph) forperforming image intensity line profile. However, the ESTEMimages shown in the figures in this paper were converted to 8-bitdisplays with a similar B/C setting to provide a reasonable demon-stration of the gases effect.

For the ESTEM images of the gas background and amorphouscarbon, as both are relatively uniform, applying the line profilingand extracting raw count intensity is straightforward (Figure 1).However, for the atomically resolved ESTEM images, line profilemeasurement could introduce ambiguity. For example, SNRs mea-sured by line profiling are different depending on if it is along theslow or the fast scan direction.[21] Here, we adopted a Matlab scriptusing the fast Fourier transform (FFT) magnitudes measurementdeveloped by Colin Ophus at Lawrence Berkeley National Labora-tory. This script locates the Bragg peaks on the FFT pattern of anESTEM lattice image. The sum of the Bragg peak intensities is the“useful signal” (amplitude squared), the sum of noise pixel intensi-ties is the “noise”, and the ratio of the two gives the SNR. Thismethod provides a relative SNR but the measurement is robustand highly reproducible. Notably to compare ESTEM images withdifferent pixel sampling (thus different size of k frequency space inFFT), we defined a unified k frequency cutoff by applying a softGaussian low pass filters. By summing the signals and noise withinthe same size of k-space for different magnifications, we can thencompare SNR across different reciprocal pixel size.

Acknowledgments

This research is part of the Chemical Imaging Initiative and aseed project conducted under the Laboratory Directed Research

and Development Program at Pacific Northwest National Labora-tory (PNNL). PNNL, a multiprogram national laboratory, is operat-

ed by Battelle for the Department of Energy under Contract DE-

AC05-76RLO1830. A portion of the research was performed usingthe Environmental Molecular Sciences Laboratory (EMSL), a na-

tional scientific user facility sponsored by the Department of En-ergy’s Office of Biological and Environmental Research and locat-

ed at Pacific Northwest National Laboratory. Special thanks toDr. Colin Ophus at Molecular Foundary, LBNL, for his generous

help and providing the FFT magnitude Matlab script. The authors

thank Dr. Libor Kovarik (EMSL, PNNL) for technical support onthe ESTEM, and we also thank Daniel Melzer and Dr. Maricruz

Sanchez-Sanchez at the Technical University of Munich for pro-viding the M1 catalysts.

Conflict of interest

The authors declare no conflict of interest.

ChemCatChem 2017, 9, 3478 – 3485 www.chemcatchem.org T 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim3484

Full Papers

Keywords: environmental scanning transmission electron

microscopy · gas atmosphere · gas–solid reaction · imageresolution · sample signal to noise ratio

[1] a) J. M. Thomas, P. A. Midgley, ChemCatChem 2010, 2, 783 – 798; b) R.Sharma, Micron 2012, 43, 1147 – 1155; c) P. A. Crozier, T. W. Hansen, MRSBull. 2015, 40, 38 – 45; d) E. D. Boyes, P. L. Gai, MRS Bull. 2015, 40, 600 –605; e) S. Helveg, C. F. Kisielowski, J. R. Jinschek, P. Specht, G. Yuan, H.Frei, Micron 2015, 68, 176 – 185.

[2] L. Marton, Bull. Acad. R. Belg. Cl. Sci. 1935, 21, 553 – 560.[3] a) H. G. Heide, J. Cell Biol. 1962, 13, 147 – 152; b) E. J. Gallegos, Rev. Sci.

Instrum. 1964, 35, 1123; c) H. Hashimoto, T. Naiki, Jpn. J. Appl. Phys.1968, 7, 946; d) P. R. Swann, High Voltage Microscope Studies of Environ-mental Reactions, Univ. Calif. 1972, pp. 878 – 904; e) H. M. Flower, J. Mi-crosc. 1973, 97, 171 – 190; f) N. M. Rodriguez, S. G. Oh, W. B. Downs, P.Pattabiraman, R. T. K. Baker, Rev. Sci. Instrum. 1990, 61, 1863 – 1868; g) R.Sharma, K. Weiss, M. McKelvy, W. Glaunsinger, in Fifty-Second AnnualMeeting—Microscopy Society of America/Twenty-Ninth Annual Meeting—Microbeam Analysis Society, Proceedings (Eds. : G. W. Bailey, A. J. Gar-rattReed), 1994, pp. 494 – 495; h) E. D. Boyes, P. L. Gai, Ultramicroscopy1997, 67, 219 – 232.

[4] a) H. Yoshida, S. Takeda, Phys. Rev. B 2005, 72, 195428; b) J. F. Creemer,S. Helveg, G. H. Hoveling, S. Ullmann, A. M. Molenbroek, P. M. Sarro,H. W. Zandbergen, Ultramicroscopy 2008, 108, 993 – 998; c) P. L. Gai, E. D.Boyes, Microsc. Res. Tech. 2009, 72, 153 – 164; d) T. W. Hansen, J. B.Wagner, R. E. Dunin-Borkowski, Mater. Sci. Technol. 2010, 26, 1338 –1344.

[5] a) M. Suzuki, T. Yaguchi, X. F. Zhang, Microscopy 2013, 62, 437 – 450;b) J. B. Wagner, M. Belaggia, in Controlled Atmosphere Transmission Elec-tron Microscopy: Principles and Practice (Eds. : T. W. Hansen, J. B.Wagner), Springer International Publishing, Cham, Switzerland, 2016.

[6] a) P. L. Hansen, J. B. Wagner, S. Helveg, J. R. Rostrup-Nielsen, B. S. Clau-sen, H. Topsoe, Science 2002, 295, 2053 – 2055; b) H. Yoshida, Y. Kuwau-chi, J. R. Jinschek, K. J. Sun, S. Tanaka, M. Kohyama, S. Shimada, M.Haruta, S. Takeda, Science 2012, 335, 317 – 319.

[7] a) P. Li, J. Liu, N. Nag, P. A. Crozier, Appl. Catal. A 2006, 307, 212 – 221;b) S. Zhang, P. N. Plessow, J. J. Willis, S. Dai, M. J. Xu, G. W. Graham, M.Cargnello, F. Abild-Pedersen, X. Q. Pan, Nano Lett. 2016, 16, 4528 – 4534.

[8] P. L. Gai, L. Lari, M. R. Ward, E. D. Boyes, Chem. Phys. Lett. 2014, 592,355 – 359.

[9] Y. Zhu, P. V. Sushko, E. Jensen, L. Kovarik, D. Melzer, C. Ophus, M. San-chez-Sanchez, J. Lercher, N. Browning, 2017, unpublished results.

[10] P. A. Crozier, in Scanning Transmission Electron Microscopy (Eds. : S. J.Pennycook, P. D. Nellist), Springer, New York, 2011, pp. 537 – 582.

[11] a) J. R. Jinschek, S. Helveg, Micron 2012, 43, 1156 – 1168; b) A. N. Bright,K. Yoshida, N. Tanaka, Ultramicroscopy 2013, 124, 46 – 51; c) H. L. Xin, K.Niu, D. H. Alsem, H. Zheng, Microsc. Microanal. 2013, 19, 1558 – 1568;d) M. Ek, S. P. F. Jespersen, C. D. Damsgaard, S. Helveg, Adv. Struct. Chem.Imaging 2016, 2, 4.

[12] a) M. M. J. Treacy, J. Microsc. Spectrosc. 1982, 7, 511 – 523; b) Z. W. Wang,Z. Y. Li, S. J. Park, A. Abdela, D. Tang, R. E. Palmer, Phys. Rev. B 2011, 84,073408.

[13] A. Rose, in Advances in Electronics and Electron Physics, Vol. 1 (Ed. : L.Marton), Academic Press, San Diego, CA, 1948, pp. 131 – 166.

[14] V. Ortalan, A. Uzun, B. C. Gates, N. D. Browning, Nat. Nanotechnol. 2010,5, 843 – 847.

[15] N. D. Browning, M. F. Chisholm, S. J. Pennycook, Nature 1993, 366, 143 –146.

[16] L. Reimer, Transmission Electron Microscopy, Springer, Heidelberg, 1984.[17] R. Danev, H. Okawara, N. Usuda, K. Kametani, K. Nagayama, J. Biol. Phys.

2002, 28, 627 – 635.[18] M. Linck, B. Freitag, S. Kujawa, M. Lehmann, T. Niermann, Ultramicrosco-

py 2012, 116, 13 – 23.[19] H. Yang, R. N. Rutte, L. Jones, M. Simson, R. Sagawa, H. Ryll, M. Huth,

T. J. Pennycook, M. L. H. Green, H. Soltau, Y. Kondo, B. G. Davis, P. D. Nel-list, Nat. Commun. 2016, 7, 12532.

[20] a) L. Kovarik, A. Stevens, A. Liyu, N. D. Browning, Appl. Phys. Lett. 2016,109, 164102; b) T. Dahmen, M. Engstler, C. Pauly, P. Trampert, N. deJonge, F. Mecklich, P. Slusallek, Sci. Rep. 2016, 6, 25350.

[21] Y. Zhu, C. Ophus, J. Ciston, H. Wang, Acta Mater. 2013, 61, 5646 – 5663.

Manuscript received: March 16, 2017

Revised manuscript received: May 21, 2017

Accepted manuscript online: May 24, 2017Version of record online: August 2, 2017

ChemCatChem 2017, 9, 3478 – 3485 www.chemcatchem.org T 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim3485

Full Papers